Health State Embedding Matching for Unstructured Clinical Text
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting a health state of a subject based on user input face challenges due to variability in input formats and formats, leading to reduced accuracy and inefficiency, with potential error sources from manual pre-processing and unreliable AI-based predictions.
Innovation Solution
A method using two machine learning models to derive and embed textual elements into numeric vectors, searching for the closest match within a set of pre-generated vectors representative of health states, allowing for efficient and reliable determination of health states even with incomplete or unstructured data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If extensive pre-processing of inputs is performed to handle variability in input formats, then measurement precision of health state determination is improved, but device complexity and loss of time increase
Solution Approach 1:
The patent introduces an embedding model as an intermediary that automatically transforms diverse textual inputs into standardized numeric vectors. This mediator handles the variability in input formats (different providers, settings, boundaries) without requiring manual pre-processing, thereby maintaining measurement precision while reducing device complexity and time loss.
Solution Approach 2:
The patent replaces manual mechanical pre-processing operations with an automated machine learning-based embedding model. The model automatically converts varied textual health state descriptions into numeric vectors, substituting the need for manual data cleaning and standardization, thus reducing both complexity and time while preserving accuracy.
2Measurement precision
If manual pre-processing is performed to handle input variability, then measurement precision is improved, but loss of time and productivity decrease
Solution Approach 1:
The patent replaces manual pre-processing operations with an automated embedding model that rapidly transforms diverse textual inputs into numeric vectors. This substitution eliminates time-consuming manual data cleaning while maintaining the precision needed for accurate health state determination, thereby significantly improving productivity.
Solution Approach 2:
The embedding model performs preliminary automatic standardization of diverse inputs into a unified numeric vector format before the main health state determination process. This preliminary action eliminates the need for subsequent manual pre-processing, saving time and improving overall processing efficiency while maintaining measurement precision.
3Device complexity
If conventional methods are used to determine health state from varied inputs, then device complexity is reduced, but reliability and measurement precision deteriorate
Solution Approach 1:
The patent introduces an embedding model as an intermediary layer that automatically handles input standardization. This mediator transforms varied textual inputs from different providers and settings into consistent numeric vectors, enabling reliable health state determination without requiring complex manual pre-processing pipelines, thus maintaining system simplicity while improving reliability.
4Productivity
If AI-based methods are used for health state prediction, then productivity is improved, but reliability deteriorates due to prediction unreliability
Solution Approach 1:
The patent replaces unreliable AI-based prediction methods with a more reliable approach using an embedding model for standardized representation followed by comparison-based determination. The embedding model automatically processes inputs efficiently while the subsequent numeric vector comparison provides more reliable and interpretable results, thus improving both productivity and reliability.
Data Source
AI summary
A computer-implemented method includes obtaining a set of first numeric vectors, each first numeric vector representative of a respective health state, the first numeric vectors having been created by using an embedding model; obtaining textual data, the textual data comprising information indicative of one or more health states of a subject; using a machine learning model for deriving, from the textual data, at least one textual element, wherein the textual element comprises information indicative of a state of the one or more health states of the subject; embedding the at least one textual element into a second numeric vector by using the embedding model; searching, among the set of first numeric vectors, a closest numeric vector that is closest to the second numeric vector; and determining the health state that is represented by the closest numeric vector as a health state indicated by the textual element.